Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.debian.user > #174022 > unrolled thread

new pc and swap

Back to article view | Back to linux.debian.user

Contents

  new pc and swap Pol Hallen <deben@fuckaround.org> - 2016-10-30 13:00 +0100
    Re: new pc and swap Michael Milliman <michael.e.milliman@gmail.com> - 2016-10-30 14:20 +0100
    Re: new pc and swap Markus Grunwald <markus@the-grue.de> - 2016-10-30 15:00 +0100
      Re: new pc and swap Richard Owlett <rowlett@cloud85.net> - 2016-10-30 15:20 +0100
        Re: new pc and swap Pascal Hambourg <pascal@plouf.fr.eu.org> - 2016-10-30 16:50 +0100
          Re: new pc and swap Markus Grunwald <markus@the-grue.de> - 2016-10-30 19:30 +0100
        Re: new pc and swap Richard Owlett <rowlett@cloud85.net> - 2016-10-31 10:30 +0100
          Re: new pc and swap Stefan Monnier <monnier@iro.umontreal.ca> - 2016-10-31 15:00 +0100
            Re: new pc and swap Richard Owlett <rowlett@cloud85.net> - 2016-10-31 19:30 +0100
              Re: new pc and swap Greg Wooledge <wooledg@eeg.ccf.org> - 2016-10-31 19:50 +0100
                Re: new pc and swap Pascal Hambourg <pascal@plouf.fr.eu.org> - 2016-10-31 20:20 +0100
              Re: new pc and swap <tomas@tuxteam.de> - 2016-10-31 21:00 +0100
              Re: new pc and swap "John L. Ries" <jries@salford-systems.com> - 2016-10-31 22:40 +0100
            Re: new pc and swap Pascal Hambourg <pascal@plouf.fr.eu.org> - 2016-10-31 20:00 +0100
              Re: new pc and swap Mark Fletcher <mark27q1@gmail.com> - 2016-11-01 00:10 +0100
                Re: new pc and swap Johann Spies <johann.spies@gmail.com> - 2016-11-01 07:10 +0100
                  Re: new pc and swap Richard Hector <richard@walnut.gen.nz> - 2016-11-01 09:30 +0100
                    Re: new pc and swap Pascal Hambourg <pascal@plouf.fr.eu.org> - 2016-11-01 11:00 +0100
                      Re: new pc and swap Richard Hector <richard@walnut.gen.nz> - 2016-11-02 01:50 +0100
                    Re: new pc and swap Stefan Monnier <monnier@iro.umontreal.ca> - 2016-11-01 14:10 +0100
                      Re: new pc and swap Pascal Hambourg <pascal@plouf.fr.eu.org> - 2016-11-01 15:00 +0100
                  Re: new pc and swap Ben Caradoc-Davies <ben@transient.nz> - 2016-11-01 20:00 +0100
            Re: new pc and swap Mark Fletcher <mark27q1@gmail.com> - 2016-11-01 00:00 +0100
    Re: new pc and swap Ben Caradoc-Davies <ben@transient.nz> - 2016-10-30 17:50 +0100
      Re: new pc and swap Pascal Hambourg <pascal@plouf.fr.eu.org> - 2016-10-30 18:20 +0100
    Re: new pc and swap David Christensen <dpchrist@holgerdanske.com> - 2016-10-30 19:50 +0100

Page 1 of 2  [1] 2  Next page →

#174022 — new pc and swap

Good sunday to all :-)

I bought a new notebook: i7 2.2Ghz, 8Gb ram and ssd 256Gb.

Consider that small disk, can I install debian without swap? Does swap 
still useful?

thanks for advices! :)

Pol

[toc] | [next] | [standalone]

#174029

On 10/30/2016 06:56 AM, Pol Hallen wrote:
> Good sunday to all :-)
>
> I bought a new notebook: i7 2.2Ghz, 8Gb ram and ssd 256Gb.
>
> Consider that small disk, can I install debian without swap? Does swap 
> still useful?
Yes, debian can be installed without swap.  Whether swap is useful 
depends on the usage you are planning on putting the machine to.  In 
some cases, swap is of great benefit; in others, not much use at all.  I 
load my machine up pretty heavy, and rarely use the swap space I have 
set up, and usually when I do get into swap space, I can slim down what 
I've got running and get back out of swap.  If you are planning on using 
hibernate, then swap is needed (I almost never use that; I usually just 
suspend to ram).  If you are planning on using several memory intensive 
applications all at the same time, then swap can be useful.  If not, not 
so much.

Hope this helps some.:-)
>
> thanks for advices! :)
>
> Pol
>

-- 
73's
Mike, WB5VQX

[toc] | [prev] | [next] | [standalone]

#174032

Hello Pol,

> Does swap still useful?

It migh be. Don't forget that you need some swap space if you want to
hibernate your System!

(You can configure extra hibernation space. I never managed to get
that right...)

--
Markus Grunwald

[toc] | [prev] | [next] | [standalone]

#174034

On 10/30/2016 8:52 AM, Markus Grunwald wrote:
> Hello Pol,
>
>> Does swap still useful?
>
> It migh be. Don't forget that you need some swap space if you want to
> hibernate your System!
>
> (You can configure extra hibernation space. I never managed to get
> that right...)
>

I once read that it was possible to swap to a named file, rather 
than a swap partition.
Is that possible with Jessie?

[toc] | [prev] | [next] | [standalone]

#174039

Le 30/10/2016 à 15:11, Richard Owlett a écrit :
> On 10/30/2016 8:52 AM, Markus Grunwald wrote:
>
>> It migh be. Don't forget that you need some swap space if you want to
>> hibernate your System!
>
> I once read that it was possible to swap to a named file, rather than a
> swap partition.
> Is that possible with Jessie?

Swap to a file is possible but IIRC hibernate to a file is not possible.

[toc] | [prev] | [next] | [standalone]

#174057

Servus Pascal,

> > I once read that it was possible to swap to a named file, rather than a
> > swap partition.
> > Is that possible with Jessie?

Creating (and using) a swapfile is easy, no problem in Jessie:

sudo mkswap /swap.0
sudo swapon /swap.0
grep swap /etc/fstab
/swap.0		swap    	swap    defaults        0

> Swap to a file is possible but IIRC hibernate to a file is not possible.

It is. In fact, I do it on my machine. Configuring it is not so easy,
but again: google is your friend.

Unfortunately, a few weeks ago, hibernating stopped working reliably
here. I wrote to this list a few days ago, without any response :(

So, hibernating to a swapfile (or a separate hibernation file) is
possible, but it doesn't seem to be a good idea. I would recommend a
swap partition. On my T61 with 4GB RAM and 128GB SSD, I have a 5GB
swapfile. As gparted works very reliably, misjudging the size of the
partition is only a small problem...

So I consider switching from swapfile to a swappartition.

cu
--
Markus Grunwald

[toc] | [prev] | [next] | [standalone]

#174070

On 10/30/2016 9:15 AM, Nicolas George wrote:
> Le nonidi 9 brumaire, an CCXXV, Richard Owlett a écrit :
>> I once read that it was possible to swap to a named file, rather than a swap
>> partition.
>> Is that possible with Jessie?
>
> Please check: fallocate, mkswap, swapon, and tell us the result.
>
> Regards,
>

As I don't have a functioning Debian system at the moment
     and
As https://manpages.debian.org/cgi-bin/man.cgi remains *DOWN* :{
I tried https://manned.org/pkg/debian-jessie as it _appeared_ to 
index all _current_ Debian manpages without success.
There were no hits in 
https://packages.debian.org/search?suite=all&section=all&arch=any&searchon=names&keywords=...
Fell back to the GENERIC https://manned.org/browse/search?q=...
That tells me swap files have existed in the past and they all 
effectively say "Beware. Here dragons have laird here." ;/

Does not answer the question I posed.
Markus' post indicates swap files may be unreliable for 
hibernation, my only likely use.
I've seen posts on multiple fora suggesting hibernation is the 
only use that will absolutely require swap. That routine use does 
not tend to require swap.

[toc] | [prev] | [next] | [standalone]

#174079

>>> I once read that it was possible to swap to a named file, rather
>>> than a swap partition.  Is that possible with Jessie?

Of course, it's possible.  But if you setup a system from scratch I'd
highly recommend you put "everything" into an LVM volume group so you
can then use an LVM volume for swap rather than a file (but with the
same advantages as using a file: it's easy to resize/create on-demand
at run-time).


        Stefan


PS: I put "everything" between quotes because I'm not sure I'd recommend
    to also put /boot in an LVM volume.  IIUC it can be made to work
    nowadays, but my systems still use a two-partition setup: one small
    partition for /boot, and the rest for LVM.

[toc] | [prev] | [next] | [standalone]

#174086

On 10/31/2016 8:54 AM, Stefan Monnier wrote:
>>>> I once read that it was possible to swap to a named file, rather
>>>> than a swap partition.  Is that possible with Jessie?
>
> Of course, it's possible.  But if you setup a system from scratch I'd
> highly recommend you put "everything" into an LVM volume group so you
> can then use an LVM volume for swap rather than a file (but with the
> same advantages as using a file: it's easy to resize/create on-demand
> at run-time).
>
>
>          Stefan
>
>
> PS: I put "everything" between quotes because I'm not sure I'd recommend
>      to also put /boot in an LVM volume.  IIUC it can be made to work
>      nowadays, but my systems still use a two-partition setup: one small
>      partition for /boot, and the rest for LVM.
>
>

HUH! U lost me ;/
To paraphrase Julius Caesar:
omnia disci en tres partes est:
   primary
   extended
   logical

Whither goest LVM?

I'm so lost as to not be able to ask an answerable question.
Help please ;/
TIA

P.S. a wise person admits his ignorance.

[toc] | [prev] | [next] | [standalone]

#174089

On Mon, Oct 31, 2016 at 01:27:22PM -0500, Richard Owlett wrote:
> To paraphrase Julius Caesar:
> omnia disci en tres partes est:
>   primary
>   extended
>   logical
> 
> Whither goest LVM?

LVM is a more flexible alternative to using a set of fixed-sized
partitions.

Under LVM, you create two (or more, but usually two) *real* partitions:
say, sda1 and sda1.  The first partition (sda1) is just formatted using
a straight file system, used for /boot.  The other partition (sda2)
is not given a file system; instead, it is turned into a Volume Group
(VG), which is sort of like a virtual disk.

Within the VG you have a whole bunch of "slices" or "physical extents"
(PE), each one being some fixed size like 4 MB or 16 MB.  The whole VG
therefore contains however many thousand PE make up the physical space.

When you want to actually use the disk, you create a "logical volume" (LV)
out of however many PEs you need.  If you're creating a 1 GB file system
for a disk with 4 MB PEs, then your new LV would consist of 256 PEs.
These don't have to be contiguous.  Once the LV is created, it is usable
just like a partition.  You can format it with a file system, and mount
it, and put files on it.  Or format it for swap, or whatever.

Typically you perform your LVM installation by setting up the file systems
and swap partitions (created on LVs) using minimum sizes, and leave most
of the VG unused.

Later, if you realize you need to increase the size of /usr by 8 GB,
you just increase the size of its LV (assigning some more PEs to it,
even if they're not contiguous to the original set), and then increase
the size of the file system to match.  With most unix-like file systems,
this doesn't require any down time (unmounting) at all.

[toc] | [prev] | [next] | [standalone]

#174094

Le 31/10/2016 à 19:45, Greg Wooledge a écrit :
> On Mon, Oct 31, 2016 at 01:27:22PM -0500, Richard Owlett wrote:
>>   primary
>>   extended
>>   logical
>>
>> Whither goest LVM?

1) Use GPT partitioning and get rid of this obsolete distinction between 
primary, extended and logical partition.

2) LVM can use any type of block device as a "container" (physical 
volume, or PV) : partitions, RAID arrays, encrypted volumes...

> Under LVM, you create two (or more, but usually two) *real* partitions:
> say, sda1 and sda1.

That looks like the same one partition.

>  The first partition (sda1) is just formatted using
> a straight file system, used for /boot.

As discussed in other messages in this thread, a separate partition with 
a plain filesystem for /boot is not always required.

>  The other partition (sda2)
> is not given a file system; instead, it is turned into a Volume Group
> (VG)

No, a Physical Volume (PV). Then you create a VG on top of one or 
several PV.

> Typically you perform your LVM installation by setting up the file systems
> and swap partitions (created on LVs) using minimum sizes, and leave most
> of the VG unused.

I would not be as extreme as you, and just advise to leave enough free 
space in the VG for future LV growth or creation. Some filesystems such 
as the ext family come with a fixed number of "inodes" (each file or 
directory consumes one inode) which depends on the initial size by 
default, and cannot be increased even when the filesystem size is 
extended. So a filesystem with minimal size may have too few inodes for 
its lifetime.

Unfortunately, the Debian installer does exactly the opposite in 
assisted mode : beside assigning insane sizes to the logical volumes, it 
allocates all the available space in the VG, making it harder to resize 
logical volumes later.

[toc] | [prev] | [next] | [standalone]

#174095

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On Mon, Oct 31, 2016 at 01:27:22PM -0500, Richard Owlett wrote:

[...]

> omnia disci en tres partes est:
>   primary
>   extended
>   logical
> 
> Whither goest LVM?

Either ou go with Pascal's advice and use GPT, or, if you feel more
comfortable, you go with the traditional DOS partitioning you sketched
above. Then, you need two "traditional" partitions:

 - one (small) primary for boot
 - another (the rest) either primary, or a logical within the extended
   (doesn't matter that much which; my Debian installer chose the second
   variant, I let it do its thing), which is then the "physical volume"
   for LVM, which does its own partition scheme within that.

In my case, the partition looks like:

  tomas@rasputin:~$ sudo fdisk -l /dev/sda
  [sudo] password for tomas: 
  
  Disk /dev/sda: 1000.2 GB, 1000204886016 bytes
  255 heads, 63 sectors/track, 121601 cylinders, total 1953525168 sectors
  Units = sectors of 1 * 512 = 512 bytes
  Sector size (logical/physical): 512 bytes / 4096 bytes
  I/O size (minimum/optimal): 4096 bytes / 4096 bytes
  Disk identifier: 0x000d2482
  
     Device Boot      Start         End      Blocks   Id  System
  /dev/sda1   *        2048     2099199     1048576   83  Linux
  /dev/sda2         2099200  1953525167   975712984    5  Extended
  /dev/sda5         2101248  1953525167   975711960   83  Linux

Note that /dev/sda2 isn't a real partition, but the extended, containing
the "real meat" /dev/sda5. This one is an encrypted Luks partition (in
my case), which then is subdivided by LVM into /, /usr, /var, /home.

What does that buy you? Well, you can, with enough care, resize your
partitions within LVM after the fact [1] (you can even add a second
hard disk and allot some or all of its space to one of the existing
volumes).

For me, the nicest part is that I can have all my separate partitions
encrypted in one fell swoop.

> P.S. a wise person admits his ignorance.

;-)

regards

[1] You have to take care of the file systems: before modifying the
  partition when shrinking, after when growing. Growing is typically
  painless and low risk and can be achieved on-line (i.e. while the
  file system is mounted), shrinking can typically only be done
  off-line (for ext4 this is so).

- -- t
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.12 (GNU/Linux)

iEYEARECAAYFAlgXoTwACgkQBcgs9XrR2kZ3oACfXlgb4qrZ1vVMdwuwIFTkeVpl
LRQAn3SAUObxiJX8gopZq8zPR51/A2Fs
=qm7Y
-----END PGP SIGNATURE-----

[toc] | [prev] | [next] | [standalone]

#174097

On Monday 2016-10-31 12:27, Richard Owlett wrote:

> Whither goest LVM?

I normally put it on a primary partition as if I'm using it, I've never
needed more than four partitions.  But I know of no reason why it can't
go on a logical partition.

--------------------------|
John L. Ries              |
Salford Systems           |
Phone: (619)543-8880 x107 |
or     (435)867-8885      |
--------------------------|


On Monday 2016-10-31 12:27, Richard Owlett wrote:

>Date: Mon, 31 Oct 2016 12:27:22
>From: Richard Owlett <rowlett@cloud85.net>
>To: debian-user@lists.debian.org
>Subject: Re: new pc and swap
>Resent-Date: Mon, 31 Oct 2016 18:27:47 +0000
>Resent-From: <debian-user@lists.debian.org>
>
> On 10/31/2016 8:54 AM, Stefan Monnier wrote:
>>>>> I once read that it was possible to swap to a named file, rather
>>>>> than a swap partition.  Is that possible with Jessie?
>>
>> Of course, it's possible.  But if you setup a system from scratch I'd
>> highly recommend you put "everything" into an LVM volume group so you
>> can then use an LVM volume for swap rather than a file (but with the
>> same advantages as using a file: it's easy to resize/create on-demand
>> at run-time).
>>
>>
>>         Stefan
>>
>>
>> PS: I put "everything" between quotes because I'm not sure I'd recommend
>>     to also put /boot in an LVM volume.  IIUC it can be made to work
>>     nowadays, but my systems still use a two-partition setup: one small
>>     partition for /boot, and the rest for LVM.
>>
>>
>
> HUH! U lost me ;/
> To paraphrase Julius Caesar:
> omnia disci en tres partes est:
>  primary
>  extended
>  logical
>
> Whither goest LVM?
>
> I'm so lost as to not be able to ask an answerable question.
> Help please ;/
> TIA
>
> P.S. a wise person admits his ignorance.
>
>
>
>

[toc] | [prev] | [next] | [standalone]

#174091

Le 31/10/2016 à 14:54, Stefan Monnier a écrit :
>
> Of course, it's possible.  But if you setup a system from scratch I'd
> highly recommend you put "everything" into an LVM volume group so you
> can then use an LVM volume for swap rather than a file (but with the
> same advantages as using a file: it's easy to resize/create on-demand
> at run-time).

I second that suggestion.
(am I an LVM fanboy ? Hmm, maybe)

> PS: I put "everything" between quotes because I'm not sure I'd recommend
>     to also put /boot in an LVM volume.  IIUC it can be made to work
>     nowadays, but my systems still use a two-partition setup: one small
>     partition for /boot, and the rest for LVM.

GRUB 2 can read LVM, so it can boot a system with /boot in LVM. I have 
used this kind of setup. The advantage is that you don't need a separate 
partition for /boot and that it provides all the benefits of LVM to 
/boot. I don't know about LILO nor any other bootloader.

However, if the LVM structure gets corrupted and GRUB cannot read it, 
then the boot process will fail before the boot menu. If /boot is on a 
separate partition which can be read, then the boot will fail only after 
loading and running the kernel and the initramfs. So the initramfs debug 
shell is still available to investigate and try to repair the issue 
without requiring to boot another system.

Also, with BIOS/legacy boot, putting /boot in LVM requires that GRUB's 
core image (boot sector) is installed in the disk boot sector (aka MBR, 
not in a partition boot sector) and GRUB's core image is installed in 
the "embedded area" on a DOS-style disk or in a dedicated "BIOS 
boot"/bios_grub partition on a GPT-style disk (not as a regular file in 
/boot/grub).

[toc] | [prev] | [next] | [standalone]

#174106

On Mon, Oct 31, 2016 at 07:53:31PM +0100, Pascal Hambourg wrote:
> Le 31/10/2016 à 14:54, Stefan Monnier a écrit :
> >
> >Of course, it's possible.  But if you setup a system from scratch I'd
> >highly recommend you put "everything" into an LVM volume group so you
> >can then use an LVM volume for swap rather than a file (but with the
> >same advantages as using a file: it's easy to resize/create on-demand
> >at run-time).
> 
> I second that suggestion.
> (am I an LVM fanboy ? Hmm, maybe)
> 
> >PS: I put "everything" between quotes because I'm not sure I'd recommend
> >    to also put /boot in an LVM volume.  IIUC it can be made to work
> >    nowadays, but my systems still use a two-partition setup: one small
> >    partition for /boot, and the rest for LVM.
> 
> GRUB 2 can read LVM, so it can boot a system with /boot in LVM. I have used
> this kind of setup. The advantage is that you don't need a separate
> partition for /boot and that it provides all the benefits of LVM to /boot. I
> don't know about LILO nor any other bootloader.
> 
> However, if the LVM structure gets corrupted and GRUB cannot read it, then
> the boot process will fail before the boot menu. If /boot is on a separate
> partition which can be read, then the boot will fail only after loading and
> running the kernel and the initramfs. So the initramfs debug shell is still
> available to investigate and try to repair the issue without requiring to
> boot another system.
> 
> Also, with BIOS/legacy boot, putting /boot in LVM requires that GRUB's core
> image (boot sector) is installed in the disk boot sector (aka MBR, not in a
> partition boot sector) and GRUB's core image is installed in the "embedded
> area" on a DOS-style disk or in a dedicated "BIOS boot"/bios_grub partition
> on a GPT-style disk (not as a regular file in /boot/grub).
> 
 This subthread (this mail and all the ones that led to it) have been 
most enlightening. Please ignore my previous mail asking for 
clarification -- should have read this subthread before replying.

Mark

[toc] | [prev] | [next] | [standalone]

#174109

[Multipart message — attachments visible in raw view] — view raw

On 1 November 2016 at 01:08, Mark Fletcher <mark27q1@gmail.com> wrote:

>
> > (am I an LVM fanboy ? Hmm, maybe)is better than life itself,
>
>
LVM has bitten me more than once in the past and I will not use it again.
In both situations it spanned more than one disk and one of the disks
failed - leaving you with unrecoverable data.

Regards
Johann

[toc] | [prev] | [next] | [standalone]

#174111

[Multipart message — attachments visible in raw view] — view raw

On 01/11/16 19:04, Johann Spies wrote:
> LVM has bitten me more than once in the past and I will not use it
> again.  In both situations it spanned more than one disk and one of the
> disks failed - leaving you with unrecoverable data.

I don't think I've ever used it like that, and probably wouldn't. At
least not unless the underlying devices were RAID1 or similar.

Spanning devices seems to me to be more or less the same as RAID0
(striping), and just as risky - perhaps slightly more convenient to
recover data from, but it certainly doesn't sound easy.

On the other hand, constraining myself to one 'device' (generally an
mdraid RAID1 device, often on partitions smaller than the disk) per vg,
I would rarely do without it.

Richard

[toc] | [prev] | [next] | [standalone]

#174114

Le 01/11/2016 à 09:25, Richard Hector a écrit :
> On 01/11/16 19:04, Johann Spies wrote:
>> LVM has bitten me more than once in the past and I will not use it
>> again.  In both situations it spanned more than one disk and one of the
>> disks failed - leaving you with unrecoverable data.
>
> I don't think I've ever used it like that, and probably wouldn't. At
> least not unless the underlying devices were RAID1 or similar.

Indeed it is highly recommended to use some redundancy (external with 
physical or software mdadm RAID, or internal within LVM) when spanning 
LVM over multiple disks.

> Spanning devices seems to me to be more or less the same as RAID0
> (striping), and just as risky

The default is to concatenate PVs like RAID "linear" (JBOD), although 
LVM can do striping too.

[toc] | [prev] | [next] | [standalone]

#174127

On 01/11/16 22:57, Pascal Hambourg wrote:
>> Spanning devices seems to me to be more or less the same as RAID0
>> (striping), and just as risky
> 
> The default is to concatenate PVs like RAID "linear" (JBOD), although
> LVM can do striping too.

True, but either way, removing one disk may damage a volume.

Richard

[toc] | [prev] | [next] | [standalone]

#174117

>> LVM has bitten me more than once in the past and I will not use it
>> again.  In both situations it spanned more than one disk and one of the
>> disks failed - leaving you with unrecoverable data.
> I don't think I've ever used it like that, and probably wouldn't. At
> least not unless the underlying devices were RAID1 or similar.

The only times I put several PVs in a VG is when moving data between
disks:
- pvcreate /dev/sd<new>
- vgextend <myVG> /dev/sd<new>
- pvmove /dev/sd<old>
- vgreduce <myVG> /dev/sd<old>

But indeed, this creates a temporary mirror and is hence safe w.r.t
disk failure.


        Stefan

[toc] | [prev] | [next] | [standalone]

Page 1 of 2  [1] 2  Next page →

Back to top | Article view | linux.debian.user

csiph-web